In the Runta example, wouldn't read/write/edit also need to target the runtime filesystem? Otherwise the model could edit one tree and run against another.
Also, bash is currently described to the model as running in the current working directory, which would no longer necessarily be true. Is the intent for a tool slot to own its schema/description as well as its implementation?
This is a great comment. Yes, for remote execution the tools need to be adjusted as well. That example is not great. I will be working on this over the next few days.
Very neat.
In the Runta example, wouldn't read/write/edit also need to target the runtime filesystem? Otherwise the model could edit one tree and run against another.
Also, bash is currently described to the model as running in the current working directory, which would no longer necessarily be true. Is the intent for a tool slot to own its schema/description as well as its implementation?
This is a great comment. Yes, for remote execution the tools need to be adjusted as well. That example is not great. I will be working on this over the next few days.
Hey, this looks great. Looking forward to other endpoint support !!
Thank you! I will be working on that next :)
A very useful project. ~Yoganand